Week 6 Interactive Visualizations

Making ggplot interactive

In this lecture, we briefly introduce a few useful packages for building interactive visuals.

  • plotly: Add basic interactions to the plot made with ggplot2.

  • leaflet: For interactive map visualizations.

  • gganimate: For graphs with animation. The package gifski converts the animation into GIF.

Let’s first load the necessary plotting libraries.

# install.packages(c("plotly", leaflet", "gganimate", "gifski"))
library(tidyverse)
library(plotly)
library(leaflet)
library(gganimate); library(gifski)

plotly

First, let us create a static plot based on the iris data set.

data(iris)
p1 <- ggplot(data = iris, aes(x = Sepal.Width, y = Petal.Width, color = Species)) +
  geom_point()
p1

plotly

plotly (https://plotly.com/) is a software that provides open-source APIs for interactive visuals.

  • In a wide variety of languages, including R, Python, Matlab, and JavaScript.

  • With the plotly package loaded, we can convert ggplot objects easily into interactive graphs.

An interactive scatter plot

  • Run the following code on your Console and interact with the resulting plot.
# Convert it into interactive visual by ggplotly
ggplotly(p1)

Additionally, we can call plotly’s own functions to build interactive graphics.

plot_ly(
  data = iris,
  x = ~Sepal.Width, y = ~Petal.Width, color = ~Species,
  type = "scatter", 
  mode = "markers") %>% 
  layout(
    title = "Iris data set visualization",
    xaxis = list(title = "Sepal width", ticksuffix = "cm"),
    yaxis = list(title = "Petal width", ticksuffix = "cm"))

3D scatter plots

plot_ly(
  data = iris, 
  x= ~Sepal.Length, y= ~Sepal.Width, z= ~Petal.Length, color = ~Species,
  type = "scatter3d",
  mode = "markers")

Time series plot

The data below are produced from US economic time series available from the FRED.

  • Let’s create a static plot on personal savings rate (psavert) and the number of unemployed in thousands (unemploy).
data(economics_long)
economics_long %>%
  filter(variable %in% c("psavert", "uempmed")) %>%
  ggplot(aes(x = date, y = value, color = variable)) +
  geom_line(lwd = 1.5)

Time series plot

For visuals with multiple time series, it is useful to include annotations or hover tools.

In the code below, we set the hovermode attribute to x unified.

  • A single hover label will appear, describing one point per trace, for points at the same x value as the cursor.
df <- economics_long %>%
  filter(variable %in% c("psavert", "uempmed"))
plot_ly(
  data = df, 
  x= ~date, y= ~value, color = ~variable,
  type = "scatter",
  mode = "lines") %>%
  # Unified hovermode
  layout(hovermode = "x unified")

df <- economics_long %>%
  filter(variable %in% c("psavert", "uempmed"))
plot_ly(
  data = df, 
  x= ~date, y= ~value, color = ~variable,
  type = "scatter",
  mode = "lines") %>%
  layout(hovermode = "x unified")

leaflet

leaflet (https://leafletjs.com/) is an open-source JavaScript library for building interactive maps.

  • Maps built with leaflet have rich interactivity by default, including the ability to pan, zoom, hover, click on map elements, and markers.

  • We can create such a maps by calling the leaflet() function.

    • Map tiles: A set of small square images, each of which shows a single piece of map.

    • By default, leaflet uses tiles from OneStreetMap (https://www.openstreetmap.org).

A single location

  • The following code pin-points a single location in Singapore.
leaflet() %>%
  addTiles() %>%
  addMarkers(lng = 103.8238, lat = 1.2540, popup = "Universal Studio Singapore")

Multiple locations

  • We can also specify multiple pairs of latitude/longitude coordinates.
data(quakes)
leaflet(quakes[1:20, ]) %>% # Show the first 20 lines on the map
  addTiles() %>%
  addMarkers(lng = ~long, lat = ~lat, 
             popup = ~paste0("Magnitude: ", mag))

Multiple locations

We can replace the location pins as circle markers via addCircleMarkers().

leaflet(quakes[1:20, ]) %>%
  addTiles() %>%
  addCircleMarkers(lng = ~long, lat = ~lat,
                   popup = ~paste0("Magnitude: ", mag),
                   radius = ~mag, stroke = FALSE, fillOpacity = 0.8)

gganimate

Lastly, let’s go through gganimate, which extends ggplot2 to include animation.

library(gapminder)
p2 <- ggplot(gapminder, aes(x = gdpPercap, y = lifeExp, size = pop, color = country)) +
  geom_point(show.legend = FALSE) +
  facet_wrap(~continent, nrow = 1) +
  labs(title = "Health vs Wealth, all years", x = "GDP per capita", y = "Life expectancy")
p2

Add animation

# Add animation to static plot
p3 <- p2 + 
  labs(title = "Wealth vs Health",
       subtitle = "Year: {frame_time}", x = "GDP per capita", y = "Life expectancy") +
  transition_time(year) 
animate(p3, renderer = gifski_renderer())

Code explained

The transition_time() function defines how the data relate to itself across time.

  • It also creates label variables we can pass to the title/subtitle via {frame_time}.

  • To save the animation as a GIF, use the following code.

# Save animation as a GIF
anim_save("../figures/gapminder.gif")

Final touches

p2 <- ggplot(gapminder, aes(x = gdpPercap, y = lifeExp, size = pop, color = country)) +
  geom_point(size = 2, show.legend = FALSE) +
  labs(title = "Health vs Wealth, all years", x = "GDP per capita", y = "Life expectancy") +
  scale_x_continuous(trans = "log10", labels = scales::label_dollar(accuracy = 1)) +
  theme_bw()
p3 <- p2 + 
  labs(title = "Wealth vs Health",
       subtitle = "Year: {frame_time}", x = "GDP per capita", y = "Life expectancy") +
  transition_time(year) 
animate(p3, renderer = gifski_renderer(), res = 150)

Let’s explore the evolution of life expectancy across continents and years.

p4 <- gapminder %>%
  group_by(continent, year) %>%
  summarize(lifeExp = mean(lifeExp, na.rm = TRUE)) %>%
  ggplot(aes(x = year, y = lifeExp, color = continent)) +
  geom_line(lwd = 2) +
  labs(title = "Life expectancy across time", x = "Year", y = "Life expectancy", color = "") +
  theme(legend.position = "top", legend.margin=margin())
p4

Animated line plot

Here, the transition_reveal() function is used to reveal new time frame progressively.

p5 <- p4 + 
  transition_reveal(year) 
animate(p5, renderer = gifski_renderer(), res = 150)

Resources

The field of visualization is moving fast - new work is constantly emerging; new products are constantly being released.

  • Early and fundamental books:

    • Exploratory Data Analysis by John Tukey (1977).

    • The Visual Display of Quantitative Information by Edward Tufte (2001).

    • The Fundamental Art: An Introduction to Information Graphics and Visualization by Alberto Cairo (2013).

  • Blogs dedicated to data visualization: